FaST-LMM-HPC is a High Performance Computing extension of the epistasis test module of FaST-LMM. The next branches (versions) are available:
- Master: FaST-LMM with the enhancements described in Martínez et al. (2017).
- Multi-GPU: An extension that can make use of multiples GPUs installed on the same node.
- MPI: An extension that can be executed on a cluster (using a single GPU per node).
- MPI-Multi-GPU: An extension that can be executed on a cluster (with multiple GPUs per node).
-
Clone this repository:
$ git clone https://github.com/epiproject/FaST-LMM-HPC.git
-
Select the desired branch (i.e., 'master').
$ git checkout master
-
Install the FaST-LMM dependencies described on the "Detailed Package Install Instructions" section of Fast-LMM.
-
In addition to those, install the next two packages that are required by FaST-LMM-HPC:
- pycuda
- skcuda
-
Finally, on the directory where you copied the source code of FaST-LMM-HPC, run:
$ sudo python setup.py install
You can use the included run_epi.py Python script to execute an standalone
epistasis test with N Python processes on an EPISTASIS_DATASET and an
EPISTAIS_PHENOTYPE in the following way:
$ python2 run_epi.py N EPISTASIS_DATASET EPISTASIS_PHENOTYPE
For example, to execute an epistasis test with 12 processes on the
all_chr.maf0.001.N300 dataset, available at the /datasets subdirectory, the
next command should be executed:
$ python2 run_epi.py 12 ./datasets/all_chr.maf0.001.N300 ./datasets/phenSynthFrom22.23.N300.txt
If you have any doubt, please do not hesitate to contact with:
- Héctor Martínez (martineh@uji.es)
Martínez, H., Barrachina, S., Castillo, M., Quintana-Ortí, E. S., De Argila, J. R., Farré, X., and Navarro, A. 2017. Accelerating FaST-LMM for epistasis tests. In Algorithms and Architectures for Parallel Processing: 17th International Conference, ICA3PP 2017, Helsinki, Finland, August 21-23, 2017, Proceedings, pages 548–557. Springer International Publishing.